Where are we now?


I am also interested in helping with Pyra kernel development, my "day job" is FPGA tooling so I'm fairly used to working at a low level including a tiny bit of kernel hacking in the past (not so much on the driver side though, more memory management and interrupt stuff). I can't commit any particular amount of time, I'd mostly be doing this for a bit of variety for what I'm working on but hopefully can manage a few hours a week. I have an OMAP5 dev kit on the way to play with.

It would be great if there were some tasks that weren't really on the critical path, but that would be useful to complete.

Biggest issues at the moment:

* Our TILER patches only work up until kernel 4.19. It doesn't work anymore on higher kernels, so rotation is missing there.
* TILER Rotation doesn't work together with 3D. If I understand correctly what @zmatt said, TILER doesn't free / allow access to the memory area the 3D driver needs, so 3D only works when TILER rotation is disabled. That conflict needs to be resolved

Those are the two main issues. The remaining stuff is optimization, but apart from that, everything works.
 
So the sound issues are now sorted?

Sound always worked, but the driver currently has the bug that it crashes when you set the volume too high.
A quick fix would be to limit the volume. However, the driver we're using is a mix between the normal driver and AESS inclusion, so I have the feeling that issue might be resolved if we simply use the mainline driver.

Hardware samplerate conversion doesn't work in the audio driver (this is what needs AESS to be working), but that's optimization.
 
Biggest issues at the moment:

* Our TILER patches only work up until kernel 4.19. It doesn't work anymore on higher kernels, so rotation is missing there.
* TILER Rotation doesn't work together with 3D. If I understand correctly what @zmatt said, TILER doesn't free / allow access to the memory area the 3D driver needs, so 3D only works when TILER rotation is disabled. That conflict needs to be resolved

Those are the two main issues. The remaining stuff is optimization, but apart from that, everything works.

...but kernel 4.14 will be supported till the end of 2023, so enough time to fix the tiler-problem or do we need a higher kernel version for other reason?

kernel.png
 
True, but every new release kernel adds more and more stuff on top the previous one.
So updating the TILER stuff from 4.14 to 5.5 is probably a lot more work than from 5.4 to 5.5, as more stuff has changed.

The best way would be to get most things (also the TILER code) upstream.
 
I'm not sure though if TILER is even enabled on HDMI... according to zmatt it shouldn't, but we still have a weirdly rotated screen on HDMI out, so I'm not sure.
Do you know where to find the patches, etc. to look at?
 
That one is Toastbuckets work kernel. It's not exactly the one we're currently using but an older clone with some fixes from Toastbucket, but any fixes for TILER there can be applied to any other 4.19 kernel as well.
So for working and testing, this one would be okay.

The end goal is to use the Letux Kernel with our own kernel config and that's it. But we can provide patches to Nikolaus when it's working.
 
Preview
I've been following the Pandora console and then the Pyra console for many years as they were something that really interested me. The Pandora was a complete disaster and I was hoping that the Pyra console would fair better. It has been about 6 years now since the announcement of the Pyra and I thought it was close to completion but I feel that interest in this project has been dwindling for quite some time, so close and yet so far. I have worked in the electronics manufacturing industry for about 25 years and I have never known any project to be dragged out for this long. I think it is time to face reality, the Pyra has failed just like the Pandora did. Evil Dragon has done a fantastic job but I really don't think his heart and sole is in the Pyra like it was at the very beginning.
I've also been following both projects for years. As you can see, I joined this site 15 years ago but haven't committed to purchasing anything until 2 months ago. I didn't own a Pandora and I do recognize it's drawbacks but I found it absolutely amazing that an online community was able to build a device on their own to fit the niche they wanted to fill. With the lessons of the Pandora, the Pyra will be even better and has a replaceable CPU board that seems very promising for future upgrades. This device isn't for everyone, it wasn't meant to be. But it will do many of the things I want it to do that nobody else offers in a package like this with a smattering of libre.
 
Preview
I've been following the Pandora console and then the Pyra console for many years as they were something that really interested me. The Pandora was a complete disaster and I was hoping that the Pyra console would fair better. It has been about 6 years now since the announcement of the Pyra and I thought it was close to completion but I feel that interest in this project has been dwindling for quite some time, so close and yet so far. I have worked in the electronics manufacturing industry for about 25 years and I have never known any project to be dragged out for this long. I think it is time to face reality, the Pyra has failed just like the Pandora did. Evil Dragon has done a fantastic job but I really don't think his heart and sole is in the Pyra like it was at the very beginning.

I think reality is something you need to face. The Pandora wasn't a failure. Did you own one? Because I don't see how you could say it was a failure if you did. I understand your concerns w/ how long the Pyra has taken; Even I can't help but feel like things somehow felt more organized back during Pandora development at times. But the Pyra isn't a failure until it has flopped and as long as ED intends to persevere and there are people interested in picking one up, (which I'm sure we'll see more of once he actually gets a production run made) then I don't believe we can judge that it has flopped or failed. Despite the arduous wait, this device still has the potential to sell well and fill a niche that isn't really filled by any competitors. (aside from the GPD Win)

-God Ginrai
 
That one is Toastbuckets work kernel. It's not exactly the one we're currently using but an older clone with some fixes from Toastbucket, but any fixes for TILER there can be applied to any other 4.19 kernel as well.
So for working and testing, this one would be okay.

The end goal is to use the Letux Kernel with our own kernel config and that's it. But we can provide patches to Nikolaus when it's working.

I see, just to double check, where are the most recent kernel and U-Boot (so I can see the kernel command line) where TILER is known working? https://projects.goldelico.com/p/gta04-kernel/ suggests it works for 4.19 but not 4.14 or 5.4, but it's slightly unclear whether that is TILER or software rotation.
 
Just in case you weren't aware and making a joke, confusing 'before' and 'until' in English seems quite a common mistake that Germans make (or at least ED always used to make that mistake, but he's been quite good at it recently).
There is no direct translation of "until" without explicitly mentioning a condition (like saying "once X happens" instead of "until X"), but for some reason this detail seems to be pretty unknown even to fairly skilled German English teachers as far as I can tell. Maybe people just kind of assume it is an alias for "before", as the German term "bevor" is very similar - I actually had to look this one up myself.

Thinking more about it, it is probably caused by the German term "bis" being very ambiguous based on context. It can be translated as both "before" as well as "until", whatever makes more sense - which works surprisingly well.
 
Last edited:
I see, just to double check, where are the most recent kernel and U-Boot (so I can see the kernel command line) where TILER is known working? https://projects.goldelico.com/p/gta04-kernel/ suggests it works for 4.19 but not 4.14 or 5.4, but it's slightly unclear whether that is TILER or software rotation.
4.19 is the most recent kernel where TILER currently works. The later kernels have changed some stuff (probably DRM) and TILER rotation stopped working then.
U-Boot has nothing to do with it AFAIK. It doesn´t even initialize our MIPI display.
 
4.19 is the most recent kernel where TILER currently works. The later kernels have changed some stuff (probably DRM) and TILER rotation stopped working then.
U-Boot has nothing to do with it AFAIK. It doesn´t even initialize our MIPI display.

Thanks for clarifying the kernel situation!

U-boot contains the kernel command line which has some rotation and FB-related parameters, even though it doesn't set anything up itself, but it seems like https://dev.pyra-handheld.com/kernel/pyra-uboot is what to work with.
 
U-boot contains the kernel command line which has some rotation and FB-related parameters, even though it doesn't set anything up itself, but it seems like https://dev.pyra-handheld.com/kernel/pyra-uboot is what to work with.

Ah, well, yes, but that's not really built-into U-Boot. We have a simple config line for that, I'll check that.

You can always PM aTc directly or even better: Come to #dragonbox-dev on freenode, as that's where aTc usually resides. And he builds our images, so he knows what sources he uses :)
Post automatically merged:

Aaand a quick update:
The first few cases have been coated. According to the company, the chrome looks A LOT better than before now, as they added a more matte scratch-resistant coating on top. So it should look like a dark, brushed alumininum now.
I'll pick these and black coated ones up tomorrow and will make pictures and videos of them during the weekend :)
 
Thanks for clarifying the kernel situation!

U-boot contains the kernel command line which has some rotation and FB-related parameters, even though it doesn't set anything up itself, but it seems like https://dev.pyra-handheld.com/kernel/pyra-uboot is what to work with.

This is where you can find the Letux kernel development https://projects.goldelico.com/p/gta04-kernel/
and this is the link for U-boot development https://projects.goldelico.com/p/gta04-uboot/

This is the letux mailing list for kernel development https://lists.goldelico.com/pipermail/letux-kernel/
This is the mailing list for gfx driver integration into Letux Kernel https://lists.goldelico.com/pipermail/openpvrsgx-devgroup/
 
Ah, well, yes, but that's not really built-into U-Boot. We have a simple config line for that, I'll check that.

You can always PM aTc directly or even better: Come to #dragonbox-dev on freenode, as that's where aTc usually resides. And he builds our images, so he knows what sources he uses :)
Post automatically merged:

Aaand a quick update:
The first few cases have been coated. According to the company, the chrome looks A LOT better than before now, as they added a more matte scratch-resistant coating on top. So it should look like a dark, brushed alumininum now.
I'll pick these and black coated ones up tomorrow and will make pictures and videos of them during the weekend :)
Can’t wait to see the chrome-style one! :cool:
 
Back
Top